Microsoft Research releases Skala 1.1 with 2.8 kcal/mol GMTKN55 benchmark error
Microsoft Research has shipped Skala 1.1, an updated deep-learning exchange-correlation functional for density functional theory workflows. The release reports a weighted average error of 2.8 kcal/mol on the GMTKN55 benchmark suite and was trained on roughly 2.5 times more data than the prior version. Availability now spans CP2K plus integrations into Psi4, FHI-aims, ORCA, and VASP, signaling a push to embed learned functionals across mainstream computational chemistry stacks rather than treat them as peripheral post-processing. Microsoft also frames the launch around expanded accessibility and a living benchmark to track performance over time. For drug discovery and materials teams, the practical shift is faster access to higher-accuracy DFT without rebuilding entire simulation pipelines. Detailed validation claims and integration timelines beyond CP2K should be confirmed against the full Microsoft Research post, as the available excerpts are truncated.
Microsoft Research releases Skala 1.1 with 2.8 kcal/mol GMTKN55 benchmark error
Microsoft Research released Skala 1.1, a deep-learning exchange-correlation functional trained on 2.5x more data than its predecessor. It achieves a weighted average error of 2.8 kcal/mol on GMTKN55 and is now available in CP2K with integrations into Psi4, FHI-aims, ORCA and VASP.
Key takeaway
Deep learning is now embedded in the exchange-correlation functional itself, and Skala 1.1 pairs that architectural shift with multi-code distribution and public benchmarking.
What happened
Microsoft Research released Skala 1.1, a deep-learning exchange-correlation functional trained on 2.5 times more data than its predecessor, according to the company's research blog announcement.
The update reports a weighted average error of 2.8 kcal/mol on GMTKN55 and is available in CP2K, with integrations into Psi4, FHI-aims, ORCA, and VASP highlighted as part of broader ecosystem access.
Evidence
Skala 1.1 was trained on 2.5x more data than its predecessor.
Microsoft Research AI · attributed
Microsoft Research released Skala 1.1, a deep-learning exchange-correlation functional trained on 2.5x more data than its predecessor.
Skala 1.1 achieves 2.8 kcal/mol weighted average error on GMTKN55.
Microsoft Research AI · attributed
It achieves a weighted average error of 2.8 kcal/mol on GMTKN55 and is now available in CP2K with integrations into Psi4, FHI-aims, ORCA and VASP.
The release emphasizes expanded accessibility and a living benchmark.
Microsoft Research AI · attributed
Skala 1.1, the updated deep-learning exchange-correlation functional from Microsoft Research, provides greater accuracy, expanded accessibility across the computational chemistry ecosystem, and a living benchmark
Why it matters
Researchers and builders in drug discovery and materials science gain a faster path to higher-accuracy molecular simulation through a functional that ships inside widely used DFT codes instead of requiring separate ML tooling.
Limits and uncertainties
Available source excerpts in the packet are truncated, so full validation scope and living-benchmark methodology are not fully visible here.
Integration status across Psi4, FHI-aims, ORCA, and VASP is described at announcement level; rollout details beyond CP2K availability are not specified in the packet.
Practical implications
Teams already running CP2K can evaluate Skala 1.1 as a drop-in exchange-correlation functional and compare GMTKN55-reported accuracy against their current workflows.
Computational chemistry operators should plan compatibility checks for Psi4, FHI-aims, ORCA, and VASP if their pipelines depend on those engines.
What to watch
Updates to Microsoft's living benchmark tracking Skala performance over time.
Concrete integration releases and documentation for Psi4, FHI-aims, ORCA, and VASP beyond the CP2K availability noted in the announcement.
